A research report issued by CITIC Construction Investment highlights that on August 3, Alibaba unveiled its latest flagship model, Qwen3.8-Max, featuring an impressive total of 2.4 trillion parameters and around 95 billion active parameters. Subsequently, on August 6, DeepSeek disclosed its intention to substantially raise API service prices in the foreseeable future, with exact particulars to be officially communicated at a later date. Prior to this, DeepSeek had adopted a peak-valley pricing strategy, where API prices during peak periods were double those during off-peak times. According to the analysis by CITIC Construction Investment, leading companies such as Alibaba are persistently scaling up their model sizes and bolstering Agent functionalities. DeepSeek's price increase, on the other hand, signals that domestic large-scale models are transitioning into a stage characterized by 'capability enhancement and commercial monetization.' Concurrently, the expansion of model parameter scales, a rise in the frequency of Agent invocations, and an uptick in API demand are expected to further escalate computing power consumption on the inference front. Consequently, it is advisable to maintain a vigilant watch on investment prospects within the domestic large-scale model and domestic computing power industry chains.